GPT-5.1-Codex API status: deprecated model, context, pricing, and migration to newer Codex models
Updated: Sep 17
GPT-5.1-Codex is now a deprecated API model, not the current Codex reference point. It remains useful for compatibility and migration work because OpenAI still documents its context, pricing, and Responses API behavior.
The practical question is no longer what GPT-5.1-Codex launched with, but whether an existing integration should keep it or move to a newer coding model.
··········
GPT-5.1-CODEX IS DEPRECATED, NOT ERASED
OpenAI continues to expose documentation for gpt-5.1-codex, but marks the model deprecated. Existing workloads can therefore encounter it even though it is no longer the recommended coding target.
........
Specification | GPT-5.1-Codex |
|---|---|
Lifecycle | Deprecated |
Primary role | Agentic coding |
Context window | 400,000 tokens |
Maximum output | 128,000 tokens |
API | Responses API |
Input / cached / output | $1.25 / $0.125 / $10 per 1M tokens |
........
··········
THE MODEL WAS BUILT FOR AGENTIC CODING
GPT-5.1-Codex specialized the GPT-5 generation for multi-step code changes, tool use, repository work, and coding-agent environments rather than general conversational use.
··········
CONTEXT AND OUTPUT LIMITS REMAIN LARGE
The documented model has a 400K context window and 128K maximum output. Those numbers are useful when estimating whether a legacy coding workflow can move without reducing repository or task context.
··········
PRICING IS STILL DOCUMENTED
Current model documentation lists $1.25 per million input tokens, $0.125 cached input, and $10 output. A migration should compare both capability and token economics, not model age alone.
··········
GPT-5.3-CODEX IS THE CURRENT CODING REFERENCE
GPT-5.3-Codex is currently documented as the most capable agentic coding model, with the same 400K context and 128K maximum output but a higher list price of $1.75 input and $14 output per million tokens.
··········
CODEX PRODUCT ACCESS AND API MODEL STATUS ARE SEPARATE
The Codex app can change which models are available when signing in with ChatGPT while API access follows the API catalog. A model disappearing from the Codex product does not by itself prove the same API endpoint has disappeared.
··········
MIGRATION SHOULD BE TESTED ON REAL REPOSITORIES
For a production coding agent, compare patch quality, tool-call behavior, retry rates, long-task completion, latency, and total token cost before replacing the old model globally.
··········
FOLLOW US FOR MORE
··········
DATA STUDIOS
··········
